Vous êtes sur la page 1sur 12

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Effects of strain on the optoelectronic properties of annealed InGaAs/GaAs self-assembled


quantum dots

This content has been downloaded from IOPscience. Please scroll down to see the full text.
2014 Semicond. Sci. Technol. 29 075013
(http://iopscience.iop.org/0268-1242/29/7/075013)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 130.102.42.98
This content was downloaded on 19/06/2014 at 08:42

Please note that terms and conditions apply.

Semiconductor Science and Technology


Semicond. Sci. Technol. 29 (2014) 075013 (11pp)

doi:10.1088/0268-1242/29/7/075013

Effects of strain on the optoelectronic


properties of annealed InGaAs/GaAs selfassembled quantum dots
M Yahyaoui1, K Sellami1, S Ben Radhia1, K Boujdaria1, M Chamarro2,
B Eble2, C Testelin2 and A Lematre3
1

Laboratoire de Physique des Matriaux: Structures et Proprits, Facult des Sciences de Bizerte,
Universit de Carthage, 7021 Zarzouna, Bizerte, Tunisia
2
Institut des Nanosciences de Paris, Universit Pierre et Marie Curie-Paris 6, CNRS-UMR 7588, 4 Place
Jussieu, 75005 Paris, France
3
Laboratoire de Photonique et Nanostructure, CNRS Route de Nozay, F-91460 Marcoussis, France
E-mail: kais.boujdaria@fsb.rnu.tn
Received 19 November 2013, revised 2 April 2014
Accepted for publication 8 April 2014
Published 16 May 2014
Abstract

The effect of the lattice-mismatch strain and of the charge carrier connement prole, on the
optical properties of thermally annealed self-assembled InxGa1xAs/GaAs quantum dots (QDs),
is theoretically analyzed by using a recently developed 40-band k.p model. First, to evaluate the
composition and size of QDs as a function of thermal annealing conditions, we model the In/Ga
interdiffusion by a Fickian diffusion. Second, we investigate the decrease of the strain effects on
the carrier connement potentials with annealing by solving the Schrdinger equation separately
for electrons and holes. It is clearly found that the strain strongly modies the QD potential
prole, leading to a different electron and hole energy distribution. Finally, we carry on a
comparison between theoretical calculations and photoluminescence (PL) experimental results
performed in thermal annealed samples. A good agreement is obtained for the energy blueshift
and the linewidth narrowing of the PL spectra measured on annealed QD ensemble. These results
prove the relevance of the present approach to describe the optoelectronic properties of the
nanostructures through the post-growth thermal annealing treatment.
Keywords: quantum dots, interdiffusion, strain, exciton energy
(Some gures may appear in colour only in the online journal)

1. Introduction

found in many heterostructure systems. One can cite the selfassembled InxGa1xAs QDs which have attracted an escalating
interest since their optical gap is very close to the optimum
wavelength for optical communication [4]. However, the
inherent problem associated with this method is the difcult
control of the size uniformity and the position of QDs. This can
result in a signicant broadening of the optical response and is
considered as a major difculty for applications using QD
heterostructures in high-efciency devices. To control the size
and the electronic levels of the InAs/GaAs self-assembled
QDs, several techniques have been used [59], and one of the
most important techniques is the thermal annealing. A thermal
treatment of the QDs leads to a blueshift, a narrowing of the

Semiconductor QDs have been the focus of intensive researches for the last 15 years [1]. The research of low-dimensional
systems such as QDs has been a subject driven by the need to
tailor electronic and optical properties for specic components
and applications. Several methods for the fabrication of QDs
have been reported over the last decade, including lithographybased technologies. The most effective and valuable method
for the fabrication of coherent, dislocation-free 10 nm QDs is
the strained layer epitaxy in the StranskiKrastanow mode
[2, 3]. Such self-organized transformation from a two-dimensional layer-by-layer growth to a three-dimensional mode was
0268-1242/14/075013+11$33.00

2014 IOP Publishing Ltd Printed in the UK

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

QDs optical emission as well as to a reduction in the QDs


interlevel spacing. It is well established that it is a powerful
technique to improve the quality of the optical and structural
properties by adjusting the QD size, shape and composition.
Moreover, it homogenizes the QD ensemble, reduces the PL
linewidth and enhances the full PL intensity. However, the
annealing also changes the atom composition of the QDs since
it causes diffusion of indium from the QDs into the barriers and
of gallium in the reverse direction, resulting in a decrease in the
connement potential and in an average size increase [9, 10].
This potential connement of the carriers inside the QD is also
strongly modied by the large strain present in the structure
induced by the great mismatch between the lattice constants of
the dot material and the substrate on which the QDs are grown.
Consequently, the accurate modeling of the strain prole in the
QD is an essential prerequisite for electronic structure calculations, and also for the clear understanding of the growth of
self-assembled QDs. Large efforts have been devoted to the
theoretical modeling of QD, and from the beginning, it was
clear that the lattice-mismatch-induced strain would play a
signicant role. In this framework, several approaches,
including strain-dependent effects, have been developed,
based either on atomistic pseudopotentiel models [11, 12],
tight-binding calculations [13], effective mass [14] or k.p
enveloppe wave-function approximations [1517]. More
recently, the effect of a thermal treatment on the QD connement proles and on the electronic levels, has been considered,
by coupling Hamiltonian and Fick diffusion equations. Due to
the complexity of the calculation, several approximations have
been done, either neglecting the strain effects [18, 19] or
assuming that the strain energy and the effective masses
depend linearly on indium concentration [20, 21].
The purpose of this paper is to apply a recently developed 40 band k.p model [22] to study the strain effects on
connement potentials in order to offer a richer picture of the
interdiffusion effect on the band proles and to obtain an
improved correlation with the PL experimental results. We
present modeling and simulations of the In/Ga interdiffusion
process in single layer InAs/GaAs QDs. We then calculate,
within the framework of the effective mass approximation,
the exciton energy levels. The present calculations have been
validated through an accurate set of comparisons with our
experimental PL spectra, measured on two series of QD,
annealed in two different regimes, under rapid or long thermal
treatment.

Figure 1. Schematic model for the cone truncated QD structure; R

denotes the QD radius, h represents the truncated cone height, d


stands for the wetting layer thickness and is the base angle.

or by cross-sectional scanning tunneling microscopy (XSTM)


[25], and successfully used to interpret experimental results
[26, 27]. A schematic model for the truncated cone QD structure
used in this work is given in gure 1. R, h, d and represent
respectively the truncated cone radius and height, the wetting
layer (WL) thickness and the base angle.
The inhomogeneity of the indium concentration in the QD
is of importance and has to be taken into account on the
connement potential or strain prole, then on the electronic
states. Recent works have evidenced indium composition
inhomogeneity in InAs/GaAs QD, by using XSTM [25, 28]. A
large indium composition gradient was estimated, with a high
indium concentration on the top of the QD (with x, indium
mole fraction, varying from 4060% at the bottom, to 100% at
the QD top). We note that such results have been obtained on
thick QD (with a height from 7 nm and to more than 10 nm),
giving rise to a low emission energy (about 1.05 eV). As we
will discuss later in this article, in the section dedicated to the
comparison between theory and experiments, the PL energy of
our as-grown samples is in the 1.221.3 eV range, and then
associated to thin QD. In a recent study, XSTM measurements
have shown that for thin QD (h = 23 nm) there is no segregation of indium on the QD top, but a regular indium prole
with its maximum at the QD center [29]. Based on these results
and on the x prole measured on QD grown previously by one
the authors [30], we assumed a uniform indium concentration
in our thin QD. The initial condition on the composition
function is chosen to be constant (0 < x 1) inside the QD/WL
and 0 inside the barrier layers.
The computation region is taken to be a cylinder of
40 nm radius and 80 nm of height in which the QD is considered to be embedded. These large dimensions are chosen in
order to reduce spurious errors arising from abrupt boundary
conditions, while keeping a reasonable computation time.
Similar initial conditions were used for instance by Srujan
et al [21], considering a pyramidal dot in a cuboidal box of
relatively large dimensions. Like other authors [18, 20], we

2. Theoretical model and simulations


2.1. Ga/In interdiffusion model

The gallium and indium interdiffusion simulations between the


dot and the barrier were performed considering a single InAs/
GaAs QD. An accurate estimate of the shape and size of the
QDs is very difcult to determine and often only available by
indirect means [23]. In our model, we have chosen truncated
cone, for the QD. Such QD shape has already been observed in
InAs/GaAs QD, either by scanning tunneling microscopy [24]
2

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

structure and a dissolution of the WL, as is demonstrated in


gure 2(i). Moreover, as shown by the black line in gure 2,
the change of the indium composition evidences a clear
increase of the QD size, leading to a carrier delocalization
with annealing.

consider this interdiffusion process in a continuum model


because the diffusion length scales are much larger than the
lattice constant. Besides, the diffusion coefcient, D, is
assumed to be independent of the space coordinates so that
we can model this diffusion process by using the second
Ficks equation
x ( r , t )
D 2x ( r , t ) = 0
t

2.2. Carrier confinement potentials variation

(1)

The knowledge of the carrier connement potential variation


is a prerequisite to the analysis of the electronic properties of
annealed QDs. In the absence of strain effects, the conning
potential for an electron and a hole is the position-dependent
difference of the band-energies between the GaAs barrier and
the QD material given by:

where x(r, t) denotes the position dependent mole fraction of


In in InxGa1xAs during the time t. We have neglected the
inuence of the strain on the interdiffusion, as suggested by
previous studies on InAs-based heterostructures: one can cite
the work of Gillin [31] who revisited the effect of annealing
on InGaAs/GaAs quantum wells, with different strain
amplitudes, at high temperature (9001050 C). More
recently, the analysis of annealed single InAs/GaAs QD by
XSTM led to the conclusion that the diffusion at the QD
region can be considered as isotropic, excluding that lateral or
vertical diffusion is favored [32].
Using the nite volume technique, we solved the
described equation in three dimensions by discretization in
both time and space, with Dirichlet boundary conditions in
order to obtain the composition of the annealed structure. We
assumed the Arrhenius equation for the temperature dependence of the diffusion constant:
E
D ( Ta ) = D0 exp a
KBTa

(3a)

(3b)

Veunstrained = QC Eg ( GaAs) Eg ( InGaAs)


Vhunstrained = QV Eg ( GaAs) Eg ( InGaAs)

where Qc and Qv = (1 Qc) are the conduction band (CB) and


valence band (VB) offset constant, respectively. The determination of the VB offset constant Qv has been the subject of
many theoretical studies (Qv = 0.280.36 [33, 34]) and
experimental works (Qv = 0.300.42 [35, 36]). In this framework, we have xed the ratio Qc/Qv = 66/34. Eg in 3(a), (b)
equations is the unstrained band gap. However, the strain
arising from the lattice mismatch between InxGa1xAs and
GaAs considerably affects the band proles. In a recent article, we have studied the effect of the strain on the band and on
the electronic structure of pure InAs QDs using a k.p theoretical approach [37]. We found that the strain effects are
signicant on both CBs and VBs. The CB edge in the strained
dot is shifted up to 0.951 eV from the value of the unstrained
InAs 0.697 eV, while the VB edge splits into two branches,
the heavy-hole (HH) and light-hole (LH) bands are separated
by 0.5 eV. Then our results, and other ones given in [3840],
revealed that the band structure of semiconductors is generally altered, due to the presence of the strain, which changes
the lattice constant and reduces the symmetry of the crystal. It
modies also the energy gaps, removes the degeneracy, and
modies the charge carrier connement potentials and their
wave functions [41]. Recently, Srujan et al [21] have clearly
pointed out the interest of a detailed calculation of the strain
effects on the connement potentials to obtain a deep
understanding of the interdiffusion effect and a better correlation with experimental results.
The strain-induced shift in the CB and VB can be
described by the formalism of the deformation potential theory [42]:

(2)

where Ta is the annealing temperature, Ea = 1.23 eV is the


activation energy of the interdiffusion process [20], KB is the
Boltzmann constant and D0 = 8.5 1014 m2 s1 is the preexponential factor [20].
Figure 2 shows the indium distribution over the annealed
QDs at 650 C (a)(c), 820 C (d)(f), and at 980 C (g)(i).
Different annealing times t equal to 10 s, 20 s, and 30 s were
considered for each annealing temperature. The as-grown QD
size parameters are: R = 25 nm, h = 3.5 nm and d = 0.5 nm.
The initial condition on the indium composition function is
chosen to be 0.6 inside the QD/WL and 0 inside the barrier
layers. The color scale represents the x values. The darkest red
color in each panel represents the maximum indium concentration but this maximum decreases with increasing temperature and annealing time. The black line represents the
indium composition isosurface at 50% level of the maximum
composition and gives an idea of the effective QD size.
A gradual indium diffusion from the QD to outside is
accompanied by a gallium diffusion in the opposite direction
starting at annealing temperature of 650 C (see gure 2).
During the thermal annealing, the atoms start diffusing in and
out of the dot. This process affects the peripheral areas of the
dot and, particularly the WL because of its small thickness.
For the highest annealing temperature 980 C, the indium
fraction in the center of the dot, after 30 s of annealing time,
does not exceed the 0.2 value. This means that due to this
high temperature, an intense interdiffusion occurred inside the
system, and thereby led to a large degradation of the QD

E LH = P +

Where
3

C
EC = 2ac 1 12
C11

(4a)

E HH = ( P + Q )

(4b)

1
Q 0 +
2

02 + 20 Q + 9Q2 (4c)

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

Figure 2. Cross sections of the calculated Indium concentration distribution of annealed QDs at 650 C (a)(c), 820 C (d)(f), and at 980 C

(g)(i). Different annealing times t equal to 10 s, 20 s, and 30 s were considered for each annealing temperature. The highest value of the
indium concentration in the QD is shown by the darkest red color in each panel. A black line gives the isosurface at 50% of the maximum
indium concentration. The as grown QD parameters used in these calculations are d = 0.5 nm, h = 3.5 nm, R = 25 nm and x = 0.6, respectively.

C
P = 2a V 1 12
C11

(5a)

2C12
Q = b 1 +

C11

(5b)

strained
VHH
= Vhunstrained + E HH

unstrained
h

=V

+ E LH

(6b)
(6c)

To investigate the effect of the strain in relation to


annealing, we considered in our calculations both the strained
and unstrained situations. Figure 3 shows the connement
potentials calculated for electrons and holes in unstrained
QDs (black line) for different annealing temperatures from
650 C to 980 C, considering a post-growth annealing time
t equal to 30 s. The electron and hole band prole calculations along the z symmetry axis ( = 0) are presented in upper
panels, and the perpendicular plane (z = 0) are presented in the
lower panels. Figure 3 shows also the connement potential
calculated for electron, HH and LH in strained QD (red line).
It can be seen, from the results of our numerical simulations,
that the annealing reduces the connement potential energy
inside the QD. This decrease is more pronounced for the
strained QD situation. This can be understood by considering
two effects. First, the interdiffusion process causes an overall
increase in QD gallium content, which reduces the band
offsets for electron, HH and LH. Second, the compressive
strain effect becomes more uniform and decreases in magnitude with annealing. This leads to a reduction of the band

In these expressions b is the shear deformation potential,


ac and av are the hydrostatic deformation potential for the CB
and VB, respectively. C11 and C12 are the elastic stiffness
constants. = ( a0 a ) a is the in-plane biaxial strain. a0
and a are, respectively, the lattice constants of GaAs
(5.6503 ) and InxGa1xAs. 0 is the Spinorbit splitting.
Since the change in the QD composition, under annealing,
results in the modication of the InxGa1xAs strains and band
parameters, we summarize in table 1 the material parameters
used to evaluate the carrier connement potentials as a
function of the indium mole fraction [16, 4345].
Under these considerations, the conning potential for an
electron and a hole, taking into account strain effects are
given by:
Vestrained = Veunstrained + Ec

strained
LH

(6a)
4

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

Figure 3. Calculated electron, LH and HH band proles, for QDs annealed at 650 C, 825 C, and 980 C. Black line (red line) represents the

unstrained (strained) QDs carrier band proles. Panels (a)(c): along the vertical symmetry plane ( = 0)). Panels (d)(f): along the in-plane
direction (z = 0).The annealing time was taken equal to 30 s for each QD. Same parameters as in gure 2.
Table 1. Band parameters of InxGa1xAs used to estimate the strain-induced changes of the connement potentials [16, 4345]. x(, z) is the
position dependent function of indium in the InxGa1xAs.

InxGa1xAs
Eg(eV)

1.519 1.58x ( , z ) + 0.47x ( , z )2 [43]

0(eV)

0.34 + 0.054x ( , z ) 0.15x ( , z )2 [43]

ac(eV)

6.3 + 3.8x ( , z ) + 1.3x ( , z )2 [44]


1.16 0.16x ( , z ) [45]
1.824 0.024x ( , z ) [16]
5.6503 0.405x ( , z ) [16]
11.88 3.55x ( , z ) [16]
5.38 0.854x ( , z ) [16]

av(eV)
b(eV)
a()
C11(x1011 dyn cm2)
C12(x1011 dyn cm2)

vior is the same for all the charge carriers, however for
electron and LH, the strain potential correction
V = Vstrained Vunstrained remains negative which means that,
due to the compressive strain, the connement potential for
these charge carriers is reduced and their energy band edges
are shifted up compared to the unstrained situation. As
already mentioned, the strain is responsible for removing the
VB degeneracy; hence the HH band edge is shifted up and the
LH one down, always with respect to the unstrained structure.
Through the variation of the energy difference as a function of
the temperature, we notice that although the splitting between
the HH and LH bands VHH VLH remains important, it is
signicantly decreased from 373.5 meV for T = 650 C to
about 111 meV for T = 980 C. That is why in the following
we will neglect the HH and LH mixing.

offsets for HH, so that both contributions (composition


change and strain reduction) induce a decrease in the HH
connement potential amplitude. For electron and LH, the
strain reduction favors an increase in the band offset. Nonetheless, the former process, gap renormalization induced by
interdiffusion, leads to a decrease of the band offsets for
electron with an increase in annealing temperature. Thus, both
potential proles become observably smoother and shallower
with the increasing-annealing temperature.
Figure 4 shows the calculation of the energy difference
between the maximum of the strained and unstrained carrier
connement potentials as a function of the annealing temperature. We considered in our case the electron, HH and LH
connement potentials. It is clearly shown that the strain
effect decreases when the temperature increases. This beha5

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

Luttinger parameters as a function of indium content for


strained and unstrained case. All the band parameters, used in
this work, are taken from [43] and are listed in table 2. The
effective HH mass along the z and -directions are expressed
in terms of the Luttinger parameters:
mhhz =

m0
1 22

(8a)

mhh =

m0
1 + 2

(8b)

m0 is the free electron mass.


To calculate the electron (hole) eigenenergies and
eigenstates, we use an exact numerical solution on a Fourier
Bessel basis over a large cylindrical domain described in
section IIA. Hence we express the (unknown) function that
corresponds to the single particle eigenstate n as a linear
combination of a certain number lmax mmax of linearly

Figure 4. Variation of the strain potential correction

V = Vstrained Vunstrained for electrons, LH and HH as a function of


the post-growth annealing temperature. Same parameters as in
gure 2.

independent functions nij , then we write the following:


n

( )

( )

where Veconf
( h) re( h)

where
ln

R e(h )

) sin (

m
Z e(h)

).

Z and R

are, respectively, the height and the radius of a large cylinder.


(, , z) denotes cylindrical coordinates, n is the main quantum number corresponding to S, P, D, F, symmetries for
the given values n = 0, 1, 2, 3, respectively, ln is the
lth root of the n-order Bessel function Jn; Clmn are the basis
coefcients and ln are the normalization constants. Thus, the
whole problem of nding the exact energy eigenvalues of the
stationary Schrdinger equation is reduced to the numerical
diagonalization of the resulting Hamiltonian matrix. We have
numerically solved the single particle Hamiltonians using a
matrix diagonalization approach. Such a technique has proved
itself to be a workable and reliable technique [42, 4749]. The
accuracy of the numerical results was well controlled through
the increase of the truncated (nite) basis set dimensionality.
Hence, we can reach a very good convergence criterion and
get the exact numerical solution of the energy eigenvalue
problem (ground state and excited states).
In our study, we focus our attention on the HH,
neglecting the coupling between HH and LH states on
account of the strain effects and large energy separation
between the zone center states (sub-levels) in thin QDs. In
gure 5, we plotted the variation of the electron and HH
ground state energies, denoted E0e and E0h , respectively, as a
function of the annealing temperature (the energies are
dened relative to the bottom of the connement potential). It
can be noticed that both electron and hole eigenenergies
decrease when increasing the annealing temperature. The
analysis of the curves leads to the conclusion that E0e is
reduced signicantly when considering the strain effect. This
energy variation under the effect of strain is particularly more
pronounced for high annealing temperatures where the calculations show that the ground-state energy is reduced by
about 44% compared to the unstrained case. For holes, the

To calculate the charge carrier energy states at different


annealing temperatures, we consider the single carrier (electron (e) or hole (h)) Hamiltonian which is written as follows:

( )

,m

nlm = ln exp ine(h) Jn

2.3. Electron and holes energy levels of annealed


quantum dots

He( h) = T re( h) + Veconf


( h) re( h)

max
= lmax
Clmn nlm ,
> 0, m > 0

(7)

is the carrier connement potential.

2
1
T re( h) = 2
is the kinetic energy of the
m *e( h) ( re( h))
carriers, m*e ( re ) and m*h ( rh ) are the effective masses of the
electron and hole, respectively.
Let us emphasize that in the major part of the publications on this subject, in order to obtain the dependence of the
band parameters on the indium composition x, linear interpolations or extrapolations are used, for the sake of simplicity
[1921]. However, this procedure is too rough to yield
satisfactory results, and sometimes is not recommended [46].
In a recent paper [43], we have made a systematic investigation of the band parameters of strained and unstrained
InxGa1xAs alloys and we have shown that the dependence of
the band parameters on In content is substantially nonlinear.
Indeed, we demonstrated that the band parameters variation
clearly exhibits for unstrained and strained materials the
bowing effect, i.e. the reduction of the band gap with respect
to the linearly interpolated value between the GaAs and
InAs band gaps. This shows that the electronic properties of
InxGa1xAs cannot be considered as simple averages of those
found in InAs and GaAs.
We have taken into account the changes in the structural
properties of QDs caused by interdiffusion and calculated the
expressions of the energy-gap, electron effective masses, the

( )

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

temperature, T = 2 K) and under a laser-diode excitation at


405 nm, focused onto the sample. The resultant QD-ensemble
PL was collected by a lens, dispersed by a monochromator
and detected by a LN2-cooled charge-coupled device camera.
The growing conditions being slightly different, the PL
spectrum of the as-grown sample II is a bit blue-shifted
compare to sample I (see gure 6). The PL spectrum has its
maximum at 1.227 eV and 1.295 eV, for samples I and II,
respectively.
To study the effect of a thermal annealing, in two different regimes, low and high diffusion coefcient, two
annealing treatments have been considered. First, we used a
rapid thermal annealing (RTA) technique. Several pieces of
sample I were annealed by applying a high temperature
between 800 and 875 C, for 30 s, with the face down on a
piece of GaAs wafer (to minimize As desorption). The optical
quality of the sample surface was unaltered even at the
highest annealing temperature. For sample II, several pieces
were annealed at a moderate temperature of 650 C, during
10, 15, 20 and 25 min, respectively, between two SiO2 pieces
and under inert gas atmosphere.
The PL spectra of the different annealed samples were
then measured with the previously described set-up. The

Figure 5. Calculated electron and HH ground state energy (E0e and

E0h , respectively) as a function of the post-growth annealing


temperature, with and without considering the strain. The zero in
vertical axes is taken at the bottom of the electron connement
potential for E0e and respectively at the top of the hole connement
potential for E0h . Same parameters as in gure 2.

Table 2. Band parameters of unstrained and strained InxGa1xAs, used in our calculations [43]. m0 is the free electron mass. As in table 1, x(,
z) is the position dependent function of Indium in the InxGa1xAs.

unstrained

strained

Eg

1.519 1.58x ( , z ) + 0.47x ( , z )2

1.519 1.155x ( , z ) + 0.17x ( , z )2

me(m0)

0.067 0.06x ( , z ) + 0.01x ( , z )2

0.067 0.03x ( , z ) 0.014x ( , z )2

1
2

7.03 3.22x ( , z ) + 13.88x ( , z )2

2.32 2.97x ( , z ) + 8.07x ( , z )2

7.03 + 1.1x ( , z ) + 11.71x ( , z )


2.32 0.12x ( , z ) + 6.26x ( , z )

normalized 2 K PL spectra are shown in gures 6(a) (rapid


thermal annealing) and (b) (slow thermal annealing at
650 C). We estimated the average ensemble exciton energy
from the PL maximum (see gures 7 and 8). An energy blueshift of about 150 meV is observed when sample I is annealed
at 875 C, meanwhile the blue-shift is less important and
about 40 meV when the sample II is annealed at 650 C for
25 min. A signicant narrowing of the PL linewidth is
observed for both annealings, however, the rapid thermal
annealing is slightly more effective, the linewidth being
reduced by 60% for sample I, instead of 30% for sample II.
We can also see from gure 6(a) that, when the spectrum is
sharpened (from 800 C), we can observe a shoulder at higher
energy of the PL spectra corresponding to the allowed optical
transitions from the excited states.
To check the reliability of our calculations, the energy of
the PL maximum is compared to the QD ground-state excitonic transition energies, namely Eexc, calculated at different
annealing temperatures as follows:

situation is a little bit different than for electrons, as the strain


slightly increases the ground state energy, and for high
annealing temperatures starting from 800 C, the energy difference between the strain and unstrained cases for the ground
state almost disappears and becomes insignicant.

3. Experimental results and comparison to the


theoretical calculations
We have studied two series of samples (noted I and II). Both
of them are as-grown samples and contain 30 layers of
InAs/GaAs QDs, separated by a 38 nm thick GaAs spacer,
with a mean density equal to 4 1010 cm2. The GaAs cap
layer is 50 nm thick and the whole heterostructure thickness is
1.2 m. They are similar to previously studied samples
[50, 51]. Sample I was p-modulation doped 2 nm below each
layer with a C-dopant density equal to about 2 1011 cm2
(leading to an average occupation of a single hole per QD
the inuence of the hole doping will be then neglected, the
exciton and trion emission energies being almost identical
with a few meV of difference). For the PL measurements, the
samples were placed in a superuid helium bath (at low

Eexc = Egs + E0e + E0h E00Coul

Where Egs is the strained band gap energy.


7

(9)

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

Figure 7. Sample I (a) calculated ground state connement energies

for the electron, E0e , and hole, E0h , and the gap energy for strained
quantum dots, Egs , versus the annealing temperature (b) calculated
exciton energy (black circles) and energy at the maximum of the PL
band (blue squares) versus the annealed temperature. The red
squares are the calculated dependence of the binding energy with the
annealing temperature.

Figure 6. Normalized PL spectra, measured at T = 2 K for (a) the

sample I series. The annealing temperature is given with the curve


legend. The annealing time is 30 s; (b) the sample II series. The
annealing time is given with the curve legend. The annealing
temperature is Tan = 650 C.

E00Coul is the electronhole Coulomb energy of the ground


exciton state. Since the localizing potential for the QD carriers
is much larger than the electronhole interaction, the Coulomb interaction energy EijCoul can be calculated in a perturbative approach. For xed electron and hole distributions, one
can write:
EijCoul

e2
=
40

ie ( re )

jh ( rh )

re rh

In order to get the best agreement between the measured


and calculated PL energies of the annealed QDs, the averages
of the height and of the radius are chosen at 3.3 nm and 24 nm
for sample I and 3.0 nm and 22 nm for sample II. A constant
value of the base angle was xed, equal to 12, consistent
with the planar transmission electron microscopy measurements [54]. Moreover, for a small base angle, it is reasonable
to assume that the strain is biaxial over all the dot area [47].
Note that if the base angle is not small, the authors of [55, 56]
veried by calculations based on the use of elastic continuum
theory that the strain distribution is expected to be quite different from the biaxial strain.
It is important to note that these small height values are
consistent with previous measurement of scanning tunneling
microscopy [24] or transmission electronic microscopy [30]
in similar QDs. Moreover, such a small height is in agreement

d 3red 3rh

(10)

where is the average dielectric constant of InGaAs [52] in


the annealed QD and i e ( re ) and jh ( rh ) are the electron and
hole envelope wave functions in the ith and jth energy state,
respectively. In order to evaluate this integral, we use the
matrix element method. For details of the calculation we refer
the reader to [53].
8

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

leading to a decrease of the connement energies, E0e and E0h


as a function of the annealing temperature or time. In
gures 7(b) and 8(b), we present the results for E00Coul as a
function of annealing temperature or annealing time. We
obtain estimates of E00Coul = 17 meV for sample I, and =16 meV
for sample II, for as-grown QDs. Our calculated values agree
well with previous studies using the Hartree approximation
[16]. During annealing, E00Coul is signicantly decreased. For
example, for sample II, after 25 min of annealing time, E00Coul
does not exceed the 6 meV value. This decrease is directly
related to the signicant change in the carrier-connement
potentials during the post-growth thermal annealing.
The theoretically calculated values are in very good
agreement with the experimental data for the two samples
series, conrming that the modeling of the QD annealing by
the diffusion model gives good results. It is important to note
that our approach describes more precisely the dependence of
PL peaks versus annealing temperature than previous similar
work. In our study, the typical disagreement between calculated and measured exciton energies is 34 meV for both
samples. This has to be compared with other studies showing
signicant differences between the experimental and simulated data for the annealed samples: the mean deviation is
typically 25 meV for [18] and 10 meV for [19, 21]. This can
be explained by three simple approximations included in
previous studies: (i) a linear interpolation used to determine
the dependence of the band parameters as a function of
indium composition (neglecting any bowing effect); (ii) the
isotropic effective HH mass, and (iii) the linear interpolation
of the strain-induced corrections to the connement potentials
Ec and EHH .
Finally, we have also calculated the full width at half
maximum (FWHM) of the PL band as a function of the
annealing temperature for the sample for which we have
measured the largest change of this parameter (sample I) and
compared to PL experimental data. The temperature dependence of the experimentally measured FWHM of PL spectrum for this sample is summarized in gure 9. Indeed, the
post-growth RTA treatments lead to a narrowing of the
FWHM from 87 meV of the as-grown sample to 33 meV of
the sample annealed at 875 C. It should be noted that a
similar behavior has been previously reported for InAs/GaAs
QD [57] and attributed to the spectral narrowing due to the
improved QD size homogeneity after interdiffusion as well as
the reduction of grown-in defect density around QDs [5, 58].
Figure 9 shows also a comparison with the theoretical
predictions. We dene E1 and E2 the energies corresponding
to the PL intensity equal to half of its maximum value
(E1 > E2). The calculated FWHM is then:

Figure 8. Sample II (a) calculated ground state connement energies

for the electron, E0e , and hole, E0h , and the gap energy for strained
quantum dots, Egs , versus the annealing time (b) calculated exciton
energy (black circles) and energy at the maximum of the PL band
(blue squares) for sample II, versus the annealing time. The red
squares represent theoretical results of the binding energy with the
annealing time.

with a high emission energy (1.221.3 eV) and a large connement, as in the case of QDs contained in our samples.
We proceed to correlate with our experimental PL
results. For that purpose, experimental data and numerical
results are presented in gure 7 for sample I, and in gure 8
for sample II. Part (a) contains the calculated Egs , E0e , and E0h ,
and part (b) shows E00Coul as well as experimental and theoretical Eexc energies as a function of annealing temperature
(gure 7) or annealing time (gure 8). We see that the
strained band gap energy increases with increasing annealing
temperature or the time of annealing. That is mainly related to
the diffusion of indium atoms outside of the QD volume, as
we have already discussed in section A. Moreover, during the
post-growth annealing, the QD height and radius increase,
while the connement potential decreases (see gure 3),

FWHM = E1 E2

(11)

For the as-grown sample I, the experimental values are


E1 = 1.27 eV and E2 = 1.18 eV, and correspond to the exciton
energies of QDs having the same radius R = 24 nm but for
heights equal to h1 = 2.8 nm and h2 = 3.9 nm, respectively. As
already said, the average size h0 is equal to 3.3 nm. Note that
9

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

signicant reduction in FWHM as a function of annealing


temperature can be used in enhancing device performance
where peak gain is critical, such as detector sensitivity at a
particular wavelength and laser gain.

4. Conclusion
To conclude, our simulations clearly demonstrate that both,
the annealing induced interdiffusion and the lattice-mismatch
strain lead to a decrease in electron and hole potential connement energies. However, at high temperature the annealing effect was found to be the dominant effect. Our results
have also revealed that thermal treatment and strain affect not
only the carrier connement potentials but also the energy
levels of annealed QDs leading to a different electron and
hole energy distribution. Hence, particularly in the case of
electrons, the ground state energy is reduced signicantly due
to strain. The present calculations have been validated
through an accurate set of comparisons with our experimental
PL spectra and FWHM. We think that an accurate modeling
of the strain prole is an essential prerequisite for electronic
structure calculations, and also for the better understanding of
the self-assembled QD growth. Moreover, to achieve a QD
with tailored features, for the realization of novel optoelectronic devices, the control of the annealing treatment is of
prime importance [5962]. In this framework, our theoretical
simulations offer a precise picture of the effect of interdiffusion and strain on band proles in self-assembled InAs/
GaAs QDs.

Figure 9. FWHM of PL band as a function of annealing temperature


for sample I. Comparison between theory (red circles) and
experiments (black squares). Inset shows the calculated relative size
dispersion, dened in equation (12), versus the annealing
temperature.

this assumption is quite reasonable and is consistent with


most of the experimental and theoretical work reported for
this system [6, 9, 21]. Indeed, many research groups have
observed that the relative variation of the QD diameter is
much smaller than the relative variation of its height.
We have previously said that the QD size is increased
when the temperature of RTA treatment is increased. In order
to calculate a dependence of the size distribution on the
annealing temperature, we apply the Ga/In interdiffussion
model presented in section II A and we dene an effective
QD size by taking into account an indium composition isosurface at 50% level of the maximum composition. As a
consequence, we can also dene an effective QD height as a
function of the thermal treatment. The inset of gure 9 shows
the relative size distribution as a function of the annealing
temperature, which is dened as follows:
h (T ) = ( h 2 (T ) h1 (T ) ) h 0 (T )

Acknowledgement
We want to acknowledge A Miard and E Galopin for the
samples synthesis and E Trimaille and P Senellart for the
thermal treatment.

(12)

References

where hi(T) is the QD height at the considered temperature,


associated to the as-grown QD for which the Eexc was equal
to Ei with the height hi (i = 1, 2). We found that the annealed
samples containing InAs QDs have narrower size distributions than the as-grown sample and that the relative size
distribution decreases from 33% to 10% when the annealing
temperature is increased to 875 C. Figure 9 also contains
theoretical results concerning FWHM as a function of
annealing temperature. Once h1(T) and h2(T) determinate, we
calculate the corresponding Eexc, i.e. E1(T) and E2(T) and then
FWHM (T). While the height is the key parameter, a slight
deviation between the experimental and theoretical results is
observed (gure 9), likely due to small effects related to QD
radius, indium composition or strain inhomogeneity. We
conclude that the narrowing of PL spectra is related to a
narrowing of size dispersion. It is also worth noting that this

[1] Reimann S M and Mannienen M 2002 Rev. Mod. Phys.


74 1283
[2] Lee J H, Wang Z M, Liang B L, Sablon K A, Strom N W and
Salamo G J 2006 Semicond. Sci. Technol. 21 1547
[3] Wang Z M 2008 Self-Assembled Quantum Dots (New York:
Springer)
[4] Alcalde A M and Weber G 1999 J. Appl. Phys. 85 7276
[5] Leon R, Kim Y, Jagadish C, gal M, zou J and Cockayne J H
1996 Appl. Phys. Lett. 69 1888
[6] Malik S, Roberts C, Murray R and Pate M 1997 Appl. Phys.
Lett. 71 1987
[7] Xu S J, Wang X C, Chua S J, Wang C H, Fan W J, Jiang J and
Xie X G 1998 Appl. Phys. Lett. 72 3335
[8] Fafard S, Wasilewski Z R, Allen C N, Picard D, Spanner M,
Mc Caffrey J P and Piva P G 1999 Phys. Rev. B 59 15368
[9] Fafard S and Allen C N 1999 Appl. Phys. Lett. 75 2374
[10] Leon R, Fafard S, Piva P G, Ruvimov S and Liliental-Weber Z
1998 Phys. Rev. B 58 R4262
10

Semicond. Sci. Technol. 29 (2014) 075013

M Yahyaoui et al

[34] Kim K, Kent P R C, Zunger A and Geller C B 2002 Phys. Rev.


B 66 045208
[35] Colombelli R, Piazza1 V, Badolato A, Lazzarino M, Beltram F,
Schoenfeld W and Petroff P 2000 Appl. Phys. Lett. 76 1146
[36] Brbach J, Yu Silov A, Haverkort J E M, Vleuten W V D and
Wolter J H 1999 Phys. Rev. B 59 10315
[37] Sadi I, Sellami K, Yahyaoui M, Testelin C and Boujdaria K
2011 J. Appl. Phys. 109 033703
[38] Jiang H and Singh J 1997 Phys. Rev. B 56 4696
[39] Lee S, Lazarenkova O L, Von Allmen P, Oyafuso F and
Klimeck G 2004 Phys. Rev. B 70 125307
[40] Adhikary S, Halder N, Chakrabarti S, Majumdar S, Ray S K,
Herrara M, Bonds M and Browning N D 2010 J. Cryst.
Growth 312 724
[41] Romanov A E, Waltereit P and Speck J S 2005 J. Appl. Phys.
97 043708
[42] Harrison P 2005 Quantum Wells, Wires and Dots: Theoretical
and Computational Physics of Semiconductor
Nanostructures (New York: Wiley)
[43] Yahyaoui M, Sellami K, Boujdaria K, Chamarro M and
Testelin C 2013 Semicond. Sci. Tehnnol. 28 125018
[44] Adachi S 1992 Physical Properties of IIIV Semiconductor
Compounds: InP, InAs, GaAs, GaP, InGaAs, and InGaAsP
(New York: Wiley)
[45] Maia A D B, da Silva E C F, Quivy A A, Bindilattti V,
de Aquino V M and Dias I F L 2012 J. Phys. D: Appl. Phys.
45 225104
[46] Ben Fredj A, Debbichi M and Said M 2007 Microelectron. J.
38 860
[47] Marzin J-Y and Bastard G 1994 Solid State Commun. 92 437
[48] Lelong P and Bastard G 1996 Solid State Commun. 98 819
[49] Quantum Dots: Research, Technology and Applications 2008
ed R W Knoss (New York: Nova Science)
[50] Sancho S, Chaouache M, Maaref M A, Bernardot F, Eble B,
Lematre A and Testelin C 2011 Phys. Rev. B 84 155458
[51] Bernardot F, Aubry E, Tribollet J, Testelin C, Chamarro M,
Lombez L, Braun P F, Marie X, Amand T and Grard J M
2006 Phys. Rev. B 73 085301
[52] Goldberg Y A and Schmidt N M 1999 Handbook Series on
Semiconductor Parameters vol 2 ed M Levinshtein,
S Rumyantsev and M Shur (London: World Scientic)
[53] Lelong P 1997 PhD Thesis Universit Pierre et Marie CurieParis 6
[54] Patan A et al 1998 J. Appl. Phys. 83 5529
[55] Grundmann M, Ledentsov N N, Stier O, Bimberg D,
Ustinov V M, Kopev P S and Alferov Z I 1996 Appl. Phys.
Lett. 68 979
[56] Grundmann M, Stier O and Bimberg D 1995 Phys. Rev. B 52
11969
[57] Fu L, Tan H H, McKerracher I, Wong-Leung J and Jagadish C
2006 J. Appl. Phys. 99 114517
[58] Lever P, Tan H H and Jagadish C 2004 J. Appl. Phys. 96 7544
[59] Aivaliotis P, Zibik E A, Wilson L R, Cockburn J W,
Hopkinson M and Airey R J 2007 Appl. Phys. Lett. 91
143502
[60] Fu L, Tan H H, Mc Kerracher I, Wong-Leung J, Jagadish C,
Vukmirovi N and Harrison P 2006 J. Appl. Phys. 99
114517
[61] Chen S-D, Chen Y-Y and Lee S-C 2005 Appl. Phys. Lett. 86
253104
[62] Zhang Z Y, Hogg R A, Xu B, Jin P and Wang Z G 2008 Opt.
Lett. 33 1210

[11] Wang J, Gong M, Guo G-C and He L 2012 J. Phys.: Condens.


Matter 24 475302
Wang L W and Zunger A 1999 Phys. Rev. B 59 15086
Williamson A J, Wang L W and Zunger A 2000 Phys. Rev. B
62 12963
[12] Jns K D, Hafenbrak R, Singh R, Ding F, Plumhof J D,
Rastelli A, Schmidt O G, Bester G and Michler P 2011 Phys.
Rev. Lett. 107 217402
Shumway J, Williamson A J, Zunger A, Passaseo A,
De Giorgi M, Cingolani R, Catalano M and Crozier P 2001
Phys. Rev. B 64 125302
[13] Robert C, Nestoklon M O, Pereira da Silva K, Pedesseau L,
Cornet C, Alonso M I, Goni A R, Turban P, Jancu J-M,
Even J and Durand O 2014 Appl. Phys. Lett. 104 011908
Santoprete R, Loiller B, Capaz R B, Kratzer P, Liu Q K K and
Schefer M 2003 Phys. Rev. B 68 235311
[14] Seravalli L, Minelli M, Frigeri P and Franchi S 2007 J. Appl.
Phys. 101 024313
Fonseca L R C, Jimenez J L, Leburton J P and Martin R M
1998 Phys. Rev. B 57 4017
[15] Pryor C 1998 Phys. Rev. B 57 7190
[16] Stier O, Grundmann M and Bimberg B 1999 Phys. Rev. B
59 5688
[17] Sheng W and Leburton J P 2001 Phys. Rev. B 63 161301
[18] Gunawan O, Djie H S and Ooi B S 2005 Phys. Rev. B 71
205319
[19] Triki M, Jaziri S and Bennaceur R 2012 J. Appl. Phys. 111
104304
[20] Petrov M Y, Ignatiev I V, Poltavtsev S V, Greilich A,
Bauschulte A, Yakovlev D R and Bayer M 2008 Phys. Rev.
B 78 045315
[21] Srujan M, Ghosh K, Sengupta S and Chakrabarti S 2010
J. Appl. Phys. 107 123107
[22] Sadi I, Ben Radhia S and Boujdaria K 2010 J. Appl. Phys. 107
043701
[23] Bruls D M, Vugs J W A M, Koenraad P M, Salemink H W M,
Wolter J H, Hopkinson M, Skolnick M S, Long F and
Gill S P A 2002 Appl. Phys. Lett. 81 1708
[24] Grandidier B, Niquet Y M, Legrand B, Nys J P, Stivenard D,
Grard J M and ThierryMieg V 2000 Phys. Rev. Lett.
85 1068
[25] Mlinar V, Bozkurt M, Ulloa J M, Ediger M, Bester G,
Badolato A, Koenraad P M, Warburton R J and Zunger A
2009 Phys. Rev. B 80 165425
[26] Oulton R, Finley J J, Tartakovskii A I, Mowbray D J,
Skolnick M S, Hopkinson M, Vasanelli A, Ferreira R and
Bastard G 2003 Phys. Rev. B 68 235301
[27] Preisler V, Grange T, Ferreira R, de Vaulchier L A, Guldner Y,
Teran F J, Potemski M and Lematre A 2006 Phys. Rev. B
73 075320
[28] Liu N, Tersoff J, Baklenov O, Holmes A L Jr and Shih C K
2000 Phys. Rev. Lett. 84 334
[29] Giddings A D, Keizer J G, Hara M, Hamhuis G J, Yuasa H,
Fukuzawa H and Koenraad P M 2011 Phys. Rev. B 83
205308
[30] Lematre A, Patriarche G and Glas F 2004 Appl. Phys. Lett.
85 3717
[31] Gillin W P 1999 J. Appl. Phys. 85 790
[32] Keizer J G, Henriques A B, Maia A D B, Quivy A A and
Koenraad P M 2012 Appl. Phys. Lett. 101 243113
[33] Kent P R C, Hart G L W and Zunger Alex 2002 Appl. Phys.
Lett. 81 4377

11

Vous aimerez peut-être aussi